In order to reduce the dynamic fuel consumption of multi-mode hybrid vehicles, the dynamic coordination control strategy of online operating condition classification was designed. The driving conditions were divided into smooth and aggressive conditions by the optimized K-means clustering(K-means) algorithm with improved particle swarm optimization and then the online operating condition identificaiton was realized. For the smooth conditions, the first-order Markov chain model was established to predict the vehicle speed and optimize the distribution of torque in real time. For the aggressive conditions, the dynamic coordination control program was designed to reduce the dynamic fuel consumption and improve the step change in the process of mode switching by adjusting the change rate of throttle opening. The simulation results show that the dynamic coordination control strategy of operating condition classification effectively reduces the fluctuation of engine torque and vehicle speed and improves the smoothness of driving with the overall fuel consumption decreased by 10.88% and the battery load state well maintained in the high efficiency area.